Comp Sci SQL WHERE Clause: Repetitions & Order Explained

[Crystal037]

Homework Statement

I am trying to understand how sql query selects multiple times the same value when I use a combination of condition in where.
I have 4 tables employee, project,department,works_on.
I have to find all project number that involve an employee with last name='Wong' either as a worker or as a manager of the department that controls the project.

Relevant Equations

select pnumber
from project,works_on,employee,department
where lname='Wong'
and ((dnum=dnumber and mgr_ssn=ssn) or (essn=ssn and pnumber=pno));

I expect output as 30,20,3,2,1 in a row but I am getting repetitions of these with 20,30 coming 3 times and 3,2 coming 18 times and 1 coming 16 times. I dont understand how combination of condition in where is leading to so many duplicates. Instead of using or if I divide the query into 2 and execute it one by one then I am getting only distinct values. Can you also explain the order by which it traverses through the table and selects tuples and display as to understand why am I getting those multiple values and why in that specific order

 

[DrJohn]

Repitition like that is usually due to a cartesian join, where some part of the join is not defined. So each row of one table is joined to EVERY row of another table. And I think that the OR is one of the things doing that. So read up on cartesian joins to see where it is going wrong.

If your select was on more than just Pnumber, say include Pno from works_on, or Pno from works_on and Dnumber from department you should see groupings of data showing how the cartesian join is collecting repetively data on a table by table basis.

It always helps to recover date where you specify the tables in the select's join rather than just a column name. You have not mentioned which table's dnum equals which table's dnumber, for example.

 

[DrJohn]

The query optimiser will test the number of rows it gets in lots of different joins and then use that to work out which join to do first, then collect data from that and choose which one it does next. This depends on indexes existing as well, if I remember correctly.

 

[jack action]

You could probably replace your first line with "select distinct pnumber" and get the result you expect. But that would be a bad fix.

The real fix you need is to rewrite your query with an appropriate JOIN syntax to eliminate the problem explained by @DrJohn . Here are links that explain how and why you should:

• Comma vs JOIN
• MySQL joins: ON vs. USING vs. Theta-style

 

[Maarten Havinga]

If you join on dnum = dnumber you get nr 5 times 9, since dnumber is 5 thrice and dnum as well. This all goes for the same person, so the join isn't uniquely linked.

 

[Maarten Havinga]

Joining with comma's or with the equivalent cross join takes every combination of rows once. If you then specify dnum = dnumber the amount of rows where both equal 5 becomes 9 since for both tables there are 3 such rows. But you also join on pnum = pnumber and these match on 2 and 3 where again dnum = dnumber = 5. Therefore these occurences (2 and 3) are again multiplied by 2, yielding 18 times 2 and 3.

BTW this also occurs on an inner join on the same conditions, you need to take a join condition that uniquely identifies one correct row.